3 ms·

Any thread on copilot always seems to get a lot of comments about how it makes dumb mistakes. I guess I'm used to Intellisense etc making dumb suggestions all t

Any thread on copilot always seems to get a lot of comments about how it makes dumb mistakes. I guess I'm used to Intellisense etc making dumb suggestions all the time? I don't dwell on it for a second, either the computer makes a good suggestion that was what I was about to do and I accept the suggestion, or it doesn't and I don't accept it. I'm certainly not going to sit there and stew over how dumb it was to suggest such a thing, it's a damn computer! Cognitive load is massively decreased, I don't have to push much on the working memory stack to do small utility functions, they just write themselves now (especially if I write the type signature first). I can spend a lot more of my time at the middle level of abstraction, and keep moving.

saurik 4y ago
Earlier in my career--back when I still relied on Intellisense--I never really thought of it as a "suggestion" that could be "wrong"?... it is a tab completion mechanism, similar to how you can tab complete filenames at the shell, and that is extremely useful if you don't exactly remember the name of the function/file you are looking for--hell: I use tab completion in bash as a replacement for ls in most cases!--but it isn't something where the thing it generates somehow looks right but is actually a security vulnerability.
- habitue 4y ago
  Yeah, I mean if you don't remember if it's `strlen`, `strnlen` when autocompleting, you can introduce a security vulnerability with intellisense just "tab completing" This is contrived, but in general (programmer + fuzzy memory + computer suggestion) means there's a possibility for it to be wrong, and you have to rely on the type system and reviewing it carefully to make sure there wasn't a mistake.